- From: Kay, Michael <Michael.Kay@softwareag.com>
- Date: Thu, 29 Jan 2004 18:41:41 +0100
- To: "Michael Dyck" <jmdyck@ibiblio.org>, <public-qt-comments@w3.org>
> > What? Then how does it handle something like this? > > validate strict context type(po:Address) { <zip>90952</zip> } > > After "validate strict", the lexer is in KINDTEST. If you > don't have a transition on "context", the lexer will have to > read "context" as a QName, which will lead to a (bogus) syntax error. I really think we need to distinguish specifying the language from writing a parser. The BNF here is absolutely clear, and it's up to the implementor whether a keyword like "context" is recognized by the lexer because it's in some special state, or whether the lexer simply returns a QName to the parser, which checks to see if the QName represents the string "context". I would love to see us get rid of lexical states entirely from the specification of the language, and reduce them to an appendix on "implementation hints". Michael Kay
Received on Thursday, 29 January 2004 12:42:01 UTC